Volume 3, Issue 22 August/September 2006
Simulating the Stars
Richard Klein uses supercomputers around the country to run his simulations.
There are one hundred billion stars in our own galaxy alone, yet we know very little about how they got there. Star formation has been one of the richest problems in astrophysics for decades. Recently though, UC Berkeley astronomer Richard Klein and his colleagues have learned a great deal about this mystery by watching the mathematics behind them unfold on a computer screen.
"You can gain insight into some of these problems with analytical approximation and solving equations with pencil and paper, but the types of systems we're dealing with now are so complicated that you can't fully describe the phenomena unless you do large scale numerical simulations," says Klein, an adjunct professor of astronomy at Berkeley who is also a staff scientist at Lawrence Livermore National Laboratory.
A slice through a 3-D simulation of a turbulent clump of molecular hydrogen, with the densest areas shown in red. The zoom-in shows a protostar accreting gas and creating a dense wake behind it. The simulation shows that a protostar, once formed, cannot accrete much more gas from the surrounding clump, contradicting the competitive accretion theory. (Credit: Mark Krumholz)
For the last ten years, Klein and his collaborators have developed high-resolution computer simulations that enable them to model their theories of star formation in three dimensions. Last year, Klein, former grad student Mark Krumholz, and physics and astronomy professor Chris McKee used their supercomputer simulations to knock down a commonly-held theory about star formation inside cold clouds of molecular hydrogen. In previous years, Klein and colleague Jonathan Arons of the Department of Astronomy not only proposed and modeled a novel theory about the formation of photon bubbles on the surface of neutron stars, but made headlines when they used an orbiting telescope to confirm their theory.
"Fifteen years ago, simulations like these would have taken 50,000 years to run on a single processor machine of the time," Klein says. "But with today's supercomputers, they might take a month running on a machine with several hundred processors working in parallel."
While the exponential increase in computing power has been a boon for Klein's research, the real secrets to his success are the computer programs he and his team develop, and the analytical work that accompanies the code. The simulations, he explains, are so complicated that they'd be meaningless unless you had an idea of what you were looking for.
"It's like mining in the side of a mountain," he says. "You really need to know the kind of jewel you're after before you start digging."
An image from NASA's Hubble Space Telescope of 30 Doradus, a vast region of gas and dust where stars are born.
Since the 1980s, researchers have attempted to use simulations to follow the birth of stars from their likely birthplace in dynamical turbulent clouds. The problem was that then state-of-the-art computers, and the code written for them, could only model the phenomena in two dimensions, an inherent discrepancy between the simulation and the real world. In the 1990s though, Klein noticed that applied mathematicians had developed extremely economical methods to solve similar equations having to do with fluid dynamics in three dimensions. Klein and his collaborators were the first to apply the method, called Adaptive Mesh Refinement, to problems in astrophysics.
"We now had the capability of solving equations that could go over many order of magnitudes of spatial scales all in one computer simulation," he says. "And we had a field day with that."
Klein and McKee founded the Berkeley Astrophysical Fluid Dynamics Group as a hub to bring this new breed of mathematical models to problems throughout the cosmos. Along the way, the group has advanced the understanding of how winds are generated from the hot accreting disks surrounding black holes and theories about the interactions of shock waves from supernovae with clumps of gas in the galaxy. Those collisions compress the clouds and may lead to the formation of new stars.
An image from a simulation of high mass star formation.
(courtesy the researchers)
These days, the group is focused on the development and modeling of a comprehensive theory of star formation, accounting for radiation, hydrodynamics, gravity, magnetic fields, and all other physical phenomena as they relate to the birth of stars. On one hand is the question of how low mass stars, like our own sun, form. Another very different problem surrounds the formation of stars that may be a hundred times the mass of the sun. These massive stars play an important role in the evolution of the galaxy as they explode into supernovae and spew forth the heavy elements that surround us and are part of us.
"It's amazing that out of pure thought you can sometimes write down a set of equations and use those to model the way that nature actually works," Klein says.